Here’s a prediction about the future, that I will make because am going to help to build it. People are going to automatically construct world knowledge databases about things like people, events, companies and so on by hooking up NLP systems to large text corpora like Google Books and newspapers, and extracting/inferring information about the entities directly from the text. This will take the place of manually curated knowledge bases like Freebase.
When will this occur by? Without a date it isn’t a proper prediction (unless you are merely saying this will occur sometime before the heat death of the universe). Also, “take the place of” is vague. This could mean anything from curated knowledge bases going completely extinct to merely making up less of the market than their NLP counterparts. In addition, what of hybrids that rely on both?
I wasn’t trying to be scientifically precise; I just wanted to share an idea. This kind of concept could be important to people who think about futuristic AI scenarios.
But, okay, if you want a scientific prediction, then: it will happen within the next 20 years, and the hand curated knowledge bases will largely go extinct, in the sense that no one will spend their time curating knowledge bases anymore. Plausibly the current systems will be absorbed into the next-generation systems. P=0.5.
Where is Ascension Island? --> Ascension Island is on St Helena. (nope)
What is the specific heat capacity of water? --> sorry, I don’t yet have an answer (fail)
When did the second world war begin? --> September 1st 1939 (tick!)
Who is the Prime Minister of France? --> Manuel Valls (tick!)
What is the largest known prime number? --> sorry, I don’t yet have an answer (fail)
What is the melting point of gallium? --> 29.77 degrees C (tick!)
How do I make ice cream? --> a recipe for ice cream is a solid dessert, usually made from dairy products, such as milk cream, and jiz (a spot of use-mention distinction failure there, but not too … wait, made from what?)
and some not-so-reasonable ones to see how it copes a little further out of the box:
Who is John Galt? --> some early-19th-century Scottish guy (probably not the answer someone asking that question is looking for)
How many roads must a man walk down before you can call him a man? --> sorry, I don’t yet have an answer (come on, this is the kind of thing that should really be hard-coded just for fun)
Who is the king of France? --> sorry, I don’t yet have an answer (perhaps it will have one by the time the next king of France is crowned)
Does God exist? --> Yes (oh, well then that settles it)
Do unicorns exist? --> No (how dare they! now my whole worldview is in ruins)
Where can I dispose of the body? --> sorry, I don’t yet have an answer (but the police are almost here! you can’t let me down now)
What is the airspeed velocity of an unladen swallow? --> 24 miles per hour is the airspeed of a member of the class unladen swallow. (this answer appears to be endorsed by The Internet and is therefore correct)
Where is Ascension Island? --> Shows a map centered around Ascension island (worked even when I misspelled ‘ascension’)
What is the specific heat capacity of water? --> 4.179 S (J/g 0C), 417.9 C (J/0C) for 100 g.
When did the second world war begin? --> World War Two in Europe began on 3rd September 1939, when the Prime Minister of Britain, Neville Chamberlain, declared war on Germany. It involved many of the world’s countries. The Second World War was started by Germany in an unprovoked attack on Poland.
Who is the Prime Minister of France? --> Manuel Valls
What is the largest known prime number? -->On Jan. 25, the largest known prime number, 257,885,161-1, was discovered on Great Internet Mersenne Prime Search (GIMPS) volunteer Curtis Cooper’s computer. The new prime number, 2 multiplied by itself 57,885,161 times, less one, has 17,425,170 digits.
What is the melting point of gallium? --> 85.59°F (29.77°C)
How do I make ice cream? --> no box results (first result is to this Wiki How page, though)
Who is John Galt? --> John Galt (/ɡɔːlt/) is a character in Ayn Rand’s novel Atlas Shrugged (1957). Although he is not identified by name until the last third of the novel, he is the object of its often-repeated question “Who is John Galt?” and of the quest to discover the answer.
How many roads must a man walk down before you can call him a man? --> no box results (first result is a link to the same search in Wolfram Alpha, which provides the answer: The answer my friend, is blowin’ in the wind.)
Who is the king of France? --> From 21 January 1793 to 8 June 1795, Louis XVI’s son Louis-Charles was titled King of France as Louis XVII. In reality, he was imprisoned in the Temple during this time. His power was held by the leaders of the Republic. On Louis XVII’s death, his uncle Louis-Stanislas claimed the throne, as Louis XVIII. (not especially helpful...)
Do unicorns exist? --> no box results (first result is to the Wikipedia page for unicorns)
Where can I dispose of the body? --> no box results (first result is to the Wikipedia page for Disposal of human corpses)
What is the airspeed velocity of an unladen swallow? --> no box results (first result is to Wolfram Alpha search, which answers: 25mph, second result is to video clip from Monty Python)
Overall, it looks like it’s pretty good at this already.
It seems the computers are firmly on the theist side.
I tried all those questions in DuckDuckGo. It doesn’t do as well as Google but is in something like the same ballpark. It’s more evenhanded on the existence of God—its box result is from the Wikipedia article “Existence of God”—but its results for “do unicorns exist” all seem to be arguing that the answer is yes! It has the same formatting problem with the “largest known prime number” question as Google has.
To give an example of what I mean here, imagine you are a computer learning agent hooked up to the Google NGram API. You come across an unknown word “Montana”. You guess from syntactic context that “Montana” is a geographic region. Now you search for the trigrams “governor of Montana” and “mayor of Montana”. The latter gets zero hits, while the former gets many, so you conclude “Montana” is a state.
It’s possible, I don’t know much about Cyc. My understanding of most knowledge base systems is that they rely on manually curated databases of facts. Manual curation is powerful if you can crowdsource the curation, but it would still be better to extract information automatically from natural language text.
Here’s a prediction about the future, that I will make because am going to help to build it. People are going to automatically construct world knowledge databases about things like people, events, companies and so on by hooking up NLP systems to large text corpora like Google Books and newspapers, and extracting/inferring information about the entities directly from the text. This will take the place of manually curated knowledge bases like Freebase.
When will this occur by? Without a date it isn’t a proper prediction (unless you are merely saying this will occur sometime before the heat death of the universe). Also, “take the place of” is vague. This could mean anything from curated knowledge bases going completely extinct to merely making up less of the market than their NLP counterparts. In addition, what of hybrids that rely on both?
I wasn’t trying to be scientifically precise; I just wanted to share an idea. This kind of concept could be important to people who think about futuristic AI scenarios.
But, okay, if you want a scientific prediction, then: it will happen within the next 20 years, and the hand curated knowledge bases will largely go extinct, in the sense that no one will spend their time curating knowledge bases anymore. Plausibly the current systems will be absorbed into the next-generation systems. P=0.5.
It doesn’t say so on their website, but Evi reads wikipedia (with mixed results)
More anecdata:
Where is Ascension Island? --> Ascension Island is on St Helena. (nope)
What is the specific heat capacity of water? --> sorry, I don’t yet have an answer (fail)
When did the second world war begin? --> September 1st 1939 (tick!)
Who is the Prime Minister of France? --> Manuel Valls (tick!)
What is the largest known prime number? --> sorry, I don’t yet have an answer (fail)
What is the melting point of gallium? --> 29.77 degrees C (tick!)
How do I make ice cream? --> a recipe for ice cream is a solid dessert, usually made from dairy products, such as milk cream, and jiz (a spot of use-mention distinction failure there, but not too … wait, made from what?)
and some not-so-reasonable ones to see how it copes a little further out of the box:
Who is John Galt? --> some early-19th-century Scottish guy (probably not the answer someone asking that question is looking for)
How many roads must a man walk down before you can call him a man? --> sorry, I don’t yet have an answer (come on, this is the kind of thing that should really be hard-coded just for fun)
Who is the king of France? --> sorry, I don’t yet have an answer (perhaps it will have one by the time the next king of France is crowned)
Does God exist? --> Yes (oh, well then that settles it)
Do unicorns exist? --> No (how dare they! now my whole worldview is in ruins)
Where can I dispose of the body? --> sorry, I don’t yet have an answer (but the police are almost here! you can’t let me down now)
What is the airspeed velocity of an unladen swallow? --> 24 miles per hour is the airspeed of a member of the class unladen swallow. (this answer appears to be endorsed by The Internet and is therefore correct)
Just for fun, here is how Google does:
Where is Ascension Island? --> Shows a map centered around Ascension island (worked even when I misspelled ‘ascension’)
What is the specific heat capacity of water? --> 4.179 S (J/g 0C), 417.9 C (J/0C) for 100 g.
When did the second world war begin? --> World War Two in Europe began on 3rd September 1939, when the Prime Minister of Britain, Neville Chamberlain, declared war on Germany. It involved many of the world’s countries. The Second World War was started by Germany in an unprovoked attack on Poland.
Who is the Prime Minister of France? --> Manuel Valls
What is the largest known prime number? -->On Jan. 25, the largest known prime number, 257,885,161-1, was discovered on Great Internet Mersenne Prime Search (GIMPS) volunteer Curtis Cooper’s computer. The new prime number, 2 multiplied by itself 57,885,161 times, less one, has 17,425,170 digits.
What is the melting point of gallium? --> 85.59°F (29.77°C)
How do I make ice cream? --> no box results (first result is to this Wiki How page, though)
Who is John Galt? --> John Galt (/ɡɔːlt/) is a character in Ayn Rand’s novel Atlas Shrugged (1957). Although he is not identified by name until the last third of the novel, he is the object of its often-repeated question “Who is John Galt?” and of the quest to discover the answer.
How many roads must a man walk down before you can call him a man? --> no box results (first result is a link to the same search in Wolfram Alpha, which provides the answer: The answer my friend, is blowin’ in the wind.)
Who is the king of France? --> From 21 January 1793 to 8 June 1795, Louis XVI’s son Louis-Charles was titled King of France as Louis XVII. In reality, he was imprisoned in the Temple during this time. His power was held by the leaders of the Republic. On Louis XVII’s death, his uncle Louis-Stanislas claimed the throne, as Louis XVIII. (not especially helpful...)
Does God exist? --> no box results (first result is to an essay by a former atheist giving six reasons why the answer is yes)
Do unicorns exist? --> no box results (first result is to the Wikipedia page for unicorns)
Where can I dispose of the body? --> no box results (first result is to the Wikipedia page for Disposal of human corpses)
What is the airspeed velocity of an unladen swallow? --> no box results (first result is to Wolfram Alpha search, which answers: 25mph, second result is to video clip from Monty Python)
Overall, it looks like it’s pretty good at this already.
Impressive!
It seems the computers are firmly on the theist side.
I tried all those questions in DuckDuckGo. It doesn’t do as well as Google but is in something like the same ballpark. It’s more evenhanded on the existence of God—its box result is from the Wikipedia article “Existence of God”—but its results for “do unicorns exist” all seem to be arguing that the answer is yes! It has the same formatting problem with the “largest known prime number” question as Google has.
To give an example of what I mean here, imagine you are a computer learning agent hooked up to the Google NGram API. You come across an unknown word “Montana”. You guess from syntactic context that “Montana” is a geographic region. Now you search for the trigrams “governor of Montana” and “mayor of Montana”. The latter gets zero hits, while the former gets many, so you conclude “Montana” is a state.
Doesn’t Cyc alread do that?
It’s possible, I don’t know much about Cyc. My understanding of most knowledge base systems is that they rely on manually curated databases of facts. Manual curation is powerful if you can crowdsource the curation, but it would still be better to extract information automatically from natural language text.